Facebook Open-Sources Presto Engine

[phoronix]

Phoronix: Facebook Open-Sources Presto Engine

Facebook has open-sourced some interesting in-house code in the past like Flashcache for the Linux kernel, the Folly C++ library, and the HipHop Virtual Machine. The latest open-source Linux-compatible software release coming out of Facebook is Presto, their tool for interacting with petabytes of information...

http://www.phoronix.com/vr.php?view=MTUwNjM

[dstaubsauger]

lol my first thought was "where the hell did they get Opera's source code"
http://en.wikipedia.org/wiki/Presto_%28layout_engine%29

[Apopas]

Hehehe same here

[MartinK]

Facebook is also using differential RPMs ?

[Cyber Killer]

lol my first thought was "where the hell did they get Opera's source code"
http://en.wikipedia.org/wiki/Presto_%28layout_engine%29

Ditto... The world needs more names, they're becoming too redundant :-P.

[Daktyl198]

I thought of Opera's engine as well, and got really excited because I would LOVE for Presto to be Open Sourced now that they've switched to Blink :P

[YoungManKlaus]

lol my first thought was "where the hell did they get Opera's source code"
http://en.wikipedia.org/wiki/Presto_%28layout_engine%29

same here ... oh god I'd love that

[kotakotakota]

"Interestingly this low-latency distributed query engine is implemented in Java but is able to avoid typical issues of Java code via writing optimized code and generating some of its own byte code."

... Java isn't slow. If you think Java is slow, you are either running into issues with the automatic garbage collector (valid concern), using Swing (... why???), or are referring to the time necessary to initialize the interpreter. Also, all Java is compiled into bytecode, so I don't know what you're talking about here.... Is there something different about what they did?

[nils_]

... Java isn't slow. If you think Java is slow, you are either running into issues with the automatic garbage collector (valid concern), using Swing (... why???), or are referring to the time necessary to initialize the interpreter. Also, all Java is compiled into bytecode, so I don't know what you're talking about here.... Is there something different about what they did?

Seems like Michael misunderstood something in the original announcement. They are referring to SQL query plans that are compiled into JVM bytecode and then use the JVM optimizations. And they probably did something fancy to avoid the GC pitfalls.

This is quoted from the announcement.

Presto dynamically compiles certain portions of the query plan down to byte code which lets the JVM optimize and generate native machine code. Through careful use of memory and data structures, Presto avoids typical issues of Java code related to memory allocation and garbage collection.